Pattern Recognition...hm... What does it actually mean? Sounds trivial, almost self-explanatory:Recognize a pattern. Here is the catch- ''a'' pattern. If we can recognize a pattern, then it must be that there is something familiar about it- evoking pleasant memories like corn cobs for pigs, or plum brandy (known as "sljivovica") for my ex-neighbor Toza (see Toza in The Pub, and Toza after exiting The Pub), or less pleasant-recognizing your own dear face on the photo from imperfect automatic photographic machine: brushed hair, nice smile, but, not a sign of your left ear, or godforbidden of your left eye, and of course, first thing you recognize about that lovely face is that marking nose, like your father, and his father, and, ...,like your mother). But, what makes a pattern familiar, or better yet- what makes pattern a pattern? A-ha! There must be something about its unique characteristics, so-called features. Finally,... Features- that is what we are after. Each of our known patterns have some specifics, something that distinguish it from its surroundings. It is the usual case that we perceive the information about those features, often not knowing how, but, then again, rather precise (if we forget about those poles or closet doors that have slammed us right in the nose more than once in this lifetime). So that, the story is something like- we perceive, extract the most powerful characteristics, proceed them up to the brain machine for processing, and wait for the result (someone shorter, others longer, depends on the machine). With the result we are given the great probability of being right (generally, could be some mistakes) we can classify the given input into its natural category, or class (this latter expression is more often used in Pattern Recognition). And, hence, we're classifying things with respect to their inherent attributes, that we have managed to recognize. Similar goes for the machine (non-human). In order to recognize something, it extracts important features first, conduct some calculations, ..., and et voila -input Sofia Loren, output- Julio Iglesias, all together- couldn't be worse (for both of them). But, that is only because we didn't tell the machine what features to choose. So, lets get serious for a while and state one of the most difficult problems in the area of Pattern Recognition:
The design of the feature extractor. What measurements should be made in order to obtain the most possible amount of information from a given source, i.e. given pattern? Since, teams and teams of psychologist haven't scratched the surface yet, regarding human perception, it makes the problem much more difficult for the machine. Thus, with the lack of confidence in our own ''lucky'' choice, we usually generate the largest possible number of the features that we can think of, just to be sure. But, does that really give us good results? ''Turn the page'' to learn more...